Re: [avr-chat] Crystals (was ATmega32 @ 16MHz : fuse bits ??)

[Graham Davies]

Vincent Trouilliez wrote:

> Q1: how much does [JTAG ICE MkII] cost ?

The price from Digi-Key here in the US is $299.  I think it's hard to find
them for less unless you can pick one up on eBay.  I've never seen one
there.

> Q2: since there are Mk1 clones, are there mk2 clones ?
> Or at least one in the pipeline ?

I don't know of any.  I don't think Atmel have released any information
about DebugWIRE and I haven't come across any reverse engineering of the
Atmel Mark II.  I am not up for reverse engineering myself.  Since I expect
people to respect my proprietary rights I must respect those of others.  If
someone does produce a Mark II clone, I very much doubt that it will be in
the $40 range you see for the Mark I clones.  Maybe you are "tinkering",
rather than working on a commercial product.  If so, why not restrict
yourself to AVRs with JTAG interfaces?  I agree so much with your earlier
statement about the need for on-chip debugging that I would be most unhappy
working on a system without it.

> ... one of the crystal terminals is obviously a lot closer to the
> AVR than the other, so maybe the capacitance is unbalanced
> between the input and output of the oscillator, and this keeps
> the oscillator from starting ??

I would doubt this.  The reason crystals are calibrated at 20 pF rather
than, say, 8 pF is so that you can install these load capacitors and they
swamp the parastic capacitance and keep things in balance.  Even so, balance
is not that critical.  Time again to justify my claims!  Remember my STK500
Expansion Board setup?  I did not solder in any load acapacitors for the
crystal.  So, why don't I just poke one in the holes?  Adding 20 pF on
either side does not stop the oscillator (15.36 MHz) and the oscillator
starts when I power cycle.  So, again, the AVR doesn't seem too fussy.  It's
seeing the load of my fingers as well as the capacitor.  I'm pretty sure I'm
making contact as the capacitor is new and the Expansion Board is
gold-finished.  Unfortunately, I don't have any other small-valued leaded
capacitors to try but I could jury-rig a surface mount one if someone thinks
this would be important.

> [the internal RC oscillator] is not accurate ...

I think you are right to insist on a crystal if you're using the UART.

> ...  I ordered an Atmega32-16PI ...

Thanks for the clarification.  So what's left?  Freak bad crystal - try
another one.  Dry joint, open or short - buzz out and inspect your circuit.
Bad fuse settings - try again, making absolutely sure you set the fuses as
you think you are.  Watch out for the programmed/unprogrammed one/zero
pitfall.  With a JTAG programmer, you can read back the fuses, even if you
stop the main clock.  How about injecting the clock that you used to rescue
your MCU and then reading back the fuses?

Graham.